Barriers to Infection of Human Cells by Feline Leukemia Virus: Insights into Resistance to Zoonosis

ABSTRACT The human genome displays a rich fossil record of past gammaretrovirus infections, yet no current epidemic is evident, despite environmental exposure to viruses that infect human cells in vitro. Feline leukemia viruses (FeLVs) rank high on this list, but neither domestic nor workplace exposure has been associated with detectable serological responses. Nonspecific inactivation of gammaretroviruses by serum factors appears insufficient to explain these observations. To investigate further, we explored the susceptibilities of primary and established human cell lines to FeLV-B, the most likely zoonotic variant. Fully permissive infection was common in cancer-derived cell lines but was also a feature of nontransformed keratinocytes and lung fibroblasts. Cells of hematopoietic origin were generally less permissive and formed discrete groups on the basis of high or low intracellular protein expression and virion release. Potent repression was observed in primary human blood mononuclear cells and a subset of leukemia cell lines. However, the early steps of reverse transcription and integration appear to be unimpaired in nonpermissive cells. FeLV-B was subject to G→A hypermutation with a predominant APOBEC3G signature in partially permissive cells but was not mutated in permissive cells or in nonpermissive cells that block secondary viral spread. Distinct cellular barriers that protect primary human blood cells are likely to be important in protection against zoonotic infection with FeLV. IMPORTANCE Domestic exposure to gammaretroviruses such as feline leukemia viruses (FeLVs) occurs worldwide, but the basis of human resistance to infection remains incompletely understood. The potential threat is evident from the human genome sequence, which reveals many past epidemics of gammaretrovirus infection, and from recent cross-species jumps of gammaretroviruses from rodents to primates and marsupials. This study examined resistance to infection at the cellular level with the most prevalent human cell-tropic FeLV variant, FeLV-B. We found that blood cells are uniquely resistant to infection with FeLV-B due to the activity of cellular enzymes that mutate the viral genome. A second block, which appears to suppress viral gene expression after the viral genome has integrated into the host cell genome, was identified. Since cells derived from other normal human cell types are fully supportive of FeLV replication, innate resistance of blood cells could be critical in protecting against cross-species infection.

The long-standing observation of the ability of some variants of feline leukemia virus (FeLV), a feline gammaretrovirus, to replicate to high titers in human cells in vitro led to early concerns with regard to zoonotic spread (6,7). FeLVs represent important pathogens of the domestic cat that are capable of cross-species spread to endangered feline species (8,9). Despite widespread domestic contact with FeLV-infected cats, which can shed virus in saliva and other body fluids, a series of careful studies has shown no evidence of serological responses in exposed individuals (10). Human serum factors that lyse or inactivate gammaretroviruses (11,12) offer a significant obstacle to cross-species spread but appear insufficient to account fully for resistance, since sensitivity can be attenuated in viruses released by human cells after initial cell-cell spread (13).
Our recent interest in the replication of FeLV in human cells was fueled by the desire to use these agents as insertional mutagens and gene discovery tools in human cancer cells in a manner similar to that of murine leukemia viruses (MLVs) in their natural host (14,15). In support of this goal, the ability of FeLV to integrate preferentially at strong promoters and enhancers is also manifested in human cells infected in vitro (16). However, it became apparent from our initial studies that the resistance of human cells to FeLV is a multifaceted phenomenon that is likely to be relevant to cross-species transfer, motivating us to conduct a deeper study.
FeLVs are a family of viruses that frequently occur as phenotypic mixtures in infected cats (17,18). Three major subgroups, which use different receptors to enter feline host cells, have been described (17). Virtually all natural isolates contain a subgroup A FeLV component (FeLV-A) that enters via the thiamine transporter THTR1 (19). This is the major horizontally transmitted form of FeLV in cats but has been considered unlikely to be zoonotic due to its low affinity for the human receptor homologue, hTHTR1 (19). Many field isolates of FeLV consist of a mixture of FeLV-A and FeLV-B, the latter arising by recombination and acquisition of envelope gene sequences from endogenous FeLV-related sequences (20). FeLV-B isolates can replicate efficiently in at least some cultured human cells with no cytopathic effect (6,7,17). Moreover, FeLV-B strains enter via the widely expressed phosphate transporter Pit-1 and/or the related transporter Pit-2 (21). Another human-cell-tropic variant is FeLV-C, which is generated by mutation of the receptor-binding domain of Env (22), facilitating entry through the heme transporter FLVCR1 or FLVCR2 (23,24). However, FeLV-C isolates are rare in nature (ϳ1% of isolates) (17) and presumably short-lived, due to their propensity to induce rapidly fatal aplastic anemia (25). A further FeLV envelope variant that has been described is the potently immunosuppressive FeLV-T, but this is also acutely pathogenic and is stringently host-specific; its complex entry requirements include a truncated envelope protein derived from endogenous FeLV-related proviruses that is released by normal feline lymphoid cells (26,27). Taken together, these biological properties implicate FeLV-B strains as the most likely candidates for zoonotic spread.
While the study of HIV and other primate lentiviruses has identified many factors that control host range by restricting replication (28), the gammaretroviruses have been less well studied in this regard. MLVs have been shown to be highly sensitive to human APOBEC3 cytidine deaminase activity but able to evade murine APOBEC3 (29,30). The effects of human APOBEC3 activity on FeLV have not been reported, although the remarkable sequence stability of the commonly transmitted form of FeLV in the domestic cat and its low mutation rate in the infection of other felids strongly suggests that it has evolved to evade feline APOBEC activity (31,32).
This study confirms FeLV-B as the variant most likely to have zoonotic potential, since virtually all human cells are susceptible, with only limited postentry barriers to infection. The resistance of primary blood cells at a postintegration step and potent APOBEC3 induction of mutations in virions released from hematopoietic cells appear to be significant factors in limiting infectivity for human cells. The possibility that FeLV could evolve to evade these barriers cannot be discounted.

RESULTS
Human cells display marked differences in susceptibility to FeLV-B infection. The rationale for focusing on FeLV-B as the most likely zoonotic variant is presented in Fig. 1A. FeLV-B is the most commonly occurring envelope variant with human cell tropism and is formed by recombination with endogenous FeLV-related sequences (20) (Fig. 1B). To generate FeLV-B virus stocks, the prototypic molecular clone of Gardner-Arnstein FeLV-B was reconstituted by transfection, propagated in HEK293 cells, which release high-titer FeLV-B into cell supernatants (10 5 to 10 6 infectious units/ml), and titrated on feline QN10S cells (33). In initial experiments, it was noted that human cells in culture were invariably susceptible to virus infection, but with markedly different proviral copy numbers as assessed by Southern blotting or quantitative PCR. This phenomenon was explored further in a dilution/spreading assay where cells were exposed at decreasing multiplicities of infection (MOI) (from 1 to 10 Ϫ4 ) and were cultured for 2 weeks to allow virus spread before the harvest of cells and analysis of FeLV proviral DNA by Southern blot hybridization. The detection of an internal 3.7-kb genome fragment provides a quantitative estimate of FeLV DNA content, as shown in indicated 30-fold-lower levels of FeLV DNA in Reh cells (ϳ0.3 copy/cell) than in semipermissive KYO-1 cells, which displayed levels similar to those in permissive controls (ϳ10 copies/cell).
As shown in Table 1, the infection/spread assay revealed full permissiveness (phenotype 1) for the replication of FeLV-B in 7 of 11 human cancer cell lines selected from the NCI-60 cancer cell panel (https://dtp.cancer.gov/discovery_development/nci-60/ cell_list.htm). Smaller numbers of cancer cell lines were partially permissive (phenotype 2) or nonpermissive (phenotype 3), with no obvious relationship to the level of expression of receptor PIT-1 mRNA as measured by quantitative reverse transcription-PCR (qRT-PCR), which was in almost all cases higher than that for the permissive HEK293 cell control) (Fig. 1D).
While spreading phenotype 1 was observed in multiple cancer cell lines, we noted that Hs68 primary foreskin fibroblasts were more restrictive. To explore the possibility that phenotype 1 is a feature of long-established cancer cell lines, we extended our study to a panel of recently established pancreatic cancer cell lines (34) as well as other nontransformed cells, including nonimmortalized CD34 ϩ cord blood cells, lung fibroblasts (IMR90), and immortalized but nontransformed skin keratinocytes (HaCAT). Where cell numbers were limiting (e.g., for CD34 ϩ cells), FeLV DNA content was analyzed by quantitative PCR. Pancreatic cancer cell lines displayed either the fully or the partially permissive phenotype, while the partially permissive phenotype was also observed in CD34 ϩ cord blood cells. The fully permissive phenotype was observed in HaCAT and IMR90 cells, indicating that permissiveness is not simply a consequence of oncogenic transformation. It is also noteworthy that we observed no evidence of cytopathic effect or significant growth alteration in any of the FeLV-B-infected cell cultures, indicating that restricted viral spread was not due to loss of cell viability or interference with cell division. In contrast, FeLV-C variants have marked cytopathic effects in some cells (35). Human hematopoietic cells display a partially permissive or nonpermissive phenotype. The permissive spreading phenotype in multiple cancer cell lines is associated with the release of high-titer FeLV at levels comparable to those in feline fibroblast controls. To explore the basis of the restricted replication phenotypes in greater depth, we examined the expression of intracellular viral antigen and virion production in representative cell lines (KYO-1 and Reh) that were exposed to FeLV-B at MOI of Ͼ1 and cultured for 2 weeks to allow viral spread. To explore the hypothesis that hematopoietic cells are more widely resistant than other cell types, we extended this analysis to a wider panel of hematopoietic cell lines as well as primary human peripheral blood mononuclear cells (PBMCs). As shown in Fig. 2A, infected KYO-1 cells express the FeLV precursor polyprotein Pr65 Gag at levels similar to those in permissive feline fibroblast cells (AH927), while Reh cells and PBMCs produce low or undetectable levels. A similar divergence was noted when virion production was analyzed. At 14 days postinfection, partially permissive KYO-1 cells showed sustained release of virions at levels similar to those in permissive feline cells, while virion release by Reh cells and PBMCs was low to undetectable ( Fig. 2A). In a separate experiment, the virtual equivalence of virion protein release was illustrated by serial dilution of pelleted virion preparations prior to Western blot analysis (Fig. 2B). As shown in Fig. 2C and Table 2, extending this analysis to a wider panel of hematopoietic cell lines showed that 9 of 13 leukemia cell lines displayed a KYO-1-like phenotype with sustained high-level release of virions, while the other 4 resembled Reh cells, with very low to undetectable levels of virus release. The level of permissiveness showed no clear pattern with regard to the type of leukemia (B cell, T cell, myeloid, erythroleukemia) or the cytogenetic features of the leukemia cell lines. PBMCs from four separate donors and a series of four Epstein-Barr virus (EBV)-immortalized lymphoblastoid cell lines (LCLs) showed a low to nonpermissive phenotype in this assay. An assay for infectious FeLV released by these cells revealed a similar pattern. Not surprisingly, cells with low virion release did not release significant levels of infectious FeLV. However, virus released at high levels from partially permissive lines showed very marked differences in specific infectivity, which was 4-to 10 4 -fold lower than that of virus released from permissive feline cells. The CEM cell line and its derivative CEMSS (36) were the most permissive lines tested, while Raji and LAMA84 cells released the least infectious virions ( Table 2).
The block to FeLV-B replication in nonpermissive cells occurs after reverse transcription. To explore the basis of the nonpermissive phenotype in greater detail, we examined the accumulation of proviral DNA after the exposure of cells to FeLV-B at an MOI of 1. Notably, nonpermissive PBMCs accumulated levels of FeLV DNA similar to those observed in partially permissive, high-virion-release KYO-1 cells (Fig. 3A). This was not due to detection of preexisting DNA in virions, since only very low levels were detected after the 30-min adsorption period, with significant increases by 6 h. To confirm that this accumulation is a receptor-dependent process, we compared FeLV-A infection of permissive feline lymphoma cells (3201 cells) to that of Reh cells, which do not express detectable mRNA for the human receptor homologue THTR1. This control was chosen because no PIT-1-negative cell lines were identified. The lack of accumulation of proviral DNA in the receptor-negative cells supports the hypothesis that this process is the result of receptor-mediated entry. The efficient detection of newly synthesized DNA with either long terminal repeat (LTR) or envelope gene primers suggests that complete proviral sequences are present, with no bias toward "strongstop" products (37).
More-extensive analysis of postadsorption amplification of proviral DNA by quantitative PCR showed that all cells tested were initially permissive regardless of the later outcome of infection (Fig. 3B). However, comparison of the quantitative change in DNA from the 6-h time point to late infection (14 days) revealed a stark contrast between the semipermissive, high-virion-release cells (e.g., CEM, Raji, and Jurkat cells), where copy numbers increased over time, and the low-virion-output, nonpermissive cells (e.g., PBMCs and Reh cells), which showed unchanged or decreasing levels of proviral DNA. Intriguingly, EBV-immortalized cell lines mainly fell into the partially permissive category in this assay, suggesting that significant viral spread had occurred in these cells, while day-14 analysis revealed only low levels of virion release with no detectable infectivity. Again, we noted no cytopathic effect or changes in cell growth in these experiments. Integrated FeLV-B DNA persists in nonpermissive human PBMCs. Despite the lack of proviral DNA amplification, we noted that FeLV-B DNA was readily detectable in PBMCs at similar levels between 3 and 14 days after initial infection (Fig. 4A). This stable association suggested that FeLV-B DNA may be integrated into PBMCs, and this hypothesis was tested further by quantitative PCR after preamplification with a FeLV Gag primer and a consensus human Alu repeat sequence (Fig. 4B). The greatly enhanced detection of FeLV DNA after Alu-Gag preamplification indicates that much of the DNA is in an integrated form. The possibility that FeLV DNA persists in an unintegrated form was also tested by PCR for circular forms generated by ligation in the nucleus. LTR circles were barely detectable in PBMCs and Reh cells by 3 days postinfection. Positive controls for this assay were early (24-h) samples from productively infected 3201 and HEK293 cells, where single and double LTR circles were detected (Fig.  4C). Cloning and sequencing confirmed the identity of the single-LTR form (not shown).
Partially permissive cells display G-to-A hypermutation that correlates with A3G mRNA expression. To explore the hypothesis that APOBEC cytidine deaminase activity is involved in FeLV growth restriction, we analyzed proviral DNA after PCR amplification of genome fragments from infected cells. These sequences were com- pared to the index FeLV-B (pFGB) sequence for evidence of characteristic G¡A mutations that are induced during reverse transcription (38). We found evidence of extensive hypermutation of FeLV replicating in human cells at rates as high as 7 mutations/kb. This mutation rate is much higher than the G¡A mutation rate reported for FeLV replicating in cats in vivo (0.1/kb) (39), indicating that FeLV, like MLV (30), is able to evade this restriction factor only in its natural host. FeLV strains typically encode a glycosylated Gag precursor open reading frame (40), and this is intact in the Gardner-Arnstein FeLV-B molecular clone. However, FeLV, like MLV, is sensitive to heterologous human APOBEC3 activity (41). As shown in Fig. 5 with the examples of the Burkitt lymphoma cell line Raji and the pancreatic cancer cell line TKCC15LO, mutational footprints indicated that the predominant signature was GG¡GA, suggesting that APOBEC3G (A3G) is responsible (42). Moreover, expression of A3G mRNA showed a positive correlation with mutational activity when all cell lines were compared (R ϭ 0.22; P ϭ 0.01). The results are collated in Fig. 5B and are broken down according to the permissive (phenotype 1), partially permissive (phenotype 2) (high virion release), or nonpermissive (phenotype 3) (low virion release) phenotype. Not surprisingly, permissive cells showed low A3G expression and low mutational activity, while partially permissive, high-virion-release cells showed high A3G expression and mutational activity. In contrast, nonpermissive cells showed no evidence of hypermutation, indicating that a nonmutational mechanism of resistance is operative in these cells. Moreover, a lack of mutation in PBMCs was observed despite very high levels of A3G, suggesting that the proviral DNA detected was the result of initial reverse transcription rather than secondary spread. The four LCLs showed significant levels of hypermutation (Fig. 5B, center, dashed oval), supporting their grouping with the partially permissive cells on the basis of proviral DNA accumulation (Fig. 3).
Analysis of all G¡A mutations in the cell lines with significant levels of hypermutation (Fig. 5C) indicated that A3G is the major, though not the sole, effector of G¡A mutations in the FeLV genome during replication in human cells. Notably, the two cell lines with divergent signatures (Hs68 and KYO-1) displayed high levels of APOBEC3F mRNA (not shown). The expression levels of A3G mRNA detected ranged over several log units but were notably higher in cells of hematopoietic origin (Fig. 5D). The CEM cell line and it derivative CEMSS showed the lowest levels among hematopoietic cell lines, in agreement with their low mutation rates and high levels of release of relatively noninfectious FeLV virions (Table 2). High levels of A3G mRNA relative to that in the HEK293 control were also associated with restricted FeLV spread in nonhematopoietic cells, but the majority of these lines showed low expression and a fully permissive phenotype. However, as indicated in Fig. 5D, the nonpermissive phenotype (nonspreading phenotype 3 [ Table 1] and low proviral DNA accumulation [ Fig. 3]) shows no relationship with A3G expression or the mutation rate. This finding, taken together with  (Table 2) but postinfection accumulation of proviral DNA (Fig. 3), are enclosed by a dashed oval. (C) Percentage of G¡A mutations that conform to the A3G signature (42) for all cell lines in which significant levels of mutations were detected. Blue-gray bars, hematopoietic cells; yellow bars, nonhematopoietic cells. (D) Relative levels of APOBEC3G mRNA (on a log 10 scale, with the level in HEK293 cells taken as 1) for all the cell lines tested, sorted into hematopoietic (blue circles) and nonhematopoietic (green circles) cell lines. Nonpermissive cells (nonspreading, with low virion release) are represented by black circles. evidence that this phenotype arises from a postintegration block to incoming virus, suggests that this is an A3G-independent barrier.

DISCUSSION
FeLV-B strains are human-cell-tropic variants that are generated frequently during FeLV infection in vivo (20,43,44) and are therefore of prime interest with respect to the zoonotic potential of FeLV. This study demonstrated fully permissive or partially permissive infection in a broad range of cancer-derived cell lines and in nontransformed keratinocytes and lung fibroblasts. Cell lines of hematopoietic origin were generally less permissive and were assigned to two discrete groups based on high or low FeLV protein expression and virion release. Primary PBMCs and some EBVtransformed LCLs were nonpermissive with respect to these measures ( Fig. 1C and 2; Table 2). While a spectrum of phenotypes was observed, this study revealed apparently universal susceptibility of human cells to initial infection with FeLV-B, with no evidence of a significant preintegration block to infection (Fig. 4); indeed, the early steps of reverse transcription and integration appear to be unimpaired in nonpermissive cells ( Fig. 3 and 5A to C). However, at least two postintegration blocks to FeLV-B replication appear to be operative in human cells. One block is clearly mediated by the APOBEC system, as indicated by the high rates of G¡A mutations in FeLV in leukemia cell lines that release abundant viral particles with low infectivity and in nonhematopoietic cancer cell lines that display a partially permissive phenotype with respect to viral spread (e.g., Raji and TKCC15LO cells, respectively) ( Table 2; Fig. 3B and 5). The predominant mutational signature was indicative of A3G in most cases, and we noted a weak correlation between A3G mRNA expression and mutational activity across a panel of cell lines (Fig. 5). This observation does not exclude the possibility that A3G also inhibits by nonmutational mechanisms, such as inhibition of reverse transcription (45).
The profound block to FeLV mRNA and protein expression in primary PBMCs (Fig. 2) and to virus release from these cells occurred without detectable mutation, despite the expression of high levels of A3G mRNA (Fig. 5), indicating a restriction that is A3G independent. APOBEC3 proteins have also been reported to block incoming murine retroviruses, and it appears that ectopic expression of A3A in mice can do so by a nonmutational mechanism (41). However, as far as we are aware, there is no evidence that APOBEC restriction can operate on incoming viruses after integration. Potential mediators of postintegration suppression of incoming viruses include epigenetic regulators, such as those that are abundantly expressed in embryonal stem cells, and restrict the expression of exogenous and endogenous retroelements (46). The possibility that the resistance of PBMCs to FeLV reflects such a general resistance mechanism against gammaretrovirus zoonosis merits further study. In this regard, we note reports of profound restriction of xenotropic murine leukemia virus-related virus (XMRV) growth in human PBMCs, which showed evidence of replication with G¡A mutations only at extremely high levels of viral input (47).
With regard to relevance to zoonotic resistance, a key question is whether the widespread susceptibility to FeLV replication observed in human cells reflects the phenotype of the tissue of origin or a loss-of-function phenomenon due to oncogenic transformation. The strongly restrictive phenotype of PBMC cultures, which contain a rich variety of primary cell types, suggests that restriction is likely to represent an important in vivo barrier. While the relative susceptibility of leukemia cell lines may be due to loss of function, the similar phenotype observed in primary CD34 ϩ cells (Table  1) suggests that this may instead reflect an immature progenitor phenotype. Moreover, many nonhematopoietic cell lines proved to be highly permissive for transcription, virion assembly, and release of FeLV-B particles, with no evidence of APOBEC-mediated mutations. Observation of this phenotype in IMR90 nontransformed lung fibroblasts, for example, which retain the capacity for replicative senescence (48), as well as in HaCAT skin keratinocytes, shows that susceptibility to infection can occur in the absence of oncogenic transformation (Table 1). Moreover, the earliest description of productive FeLV replication in human cells was based on infection of HEL2000 primary lung fibroblasts (6). In the case of transformed cells, it is difficult to discern whether the relative susceptibility to FeLV represents a normal cellular state relevant to potential zoonosis. While very-high-level A3G expression and mutational activity were observed in the TKCC15LO pancreatic cancer cell line (Fig. 5), this may not reflect the normal phenotype of pancreatic tissue. Upregulated A3G has been postulated to play a direct oncogenic role in some pancreatic cancers (49).
The susceptibility of human cancer cells to spreading FeLV-B replication provides a potentially instructive parallel with the mobilization and frequent reinsertion of L1 retrotransposons in Ͼ50% of human cancers (50); high levels were noted in lung, colon, prostate, and breast cancers, while the data sets for other tumor types showed few if any new insertions in renal carcinomas and melanomas. These findings parallel the present observations in that colon, lung, and breast cancer cell lines were fully permissive to FeLV infection, but renal cancer cells and one melanoma line displayed a strong early block to FeLV replication, with no evidence of APOBEC-mediated mutations ( Table 1; Fig. 5B). A further intriguing parallel is provided by a recent study of XMRV pathogenesis in macaques, where intravenous inoculation of high-titer virus led to the dissemination and persistence of viral protein expression, with the highest levels in lung, colon, and prostate tissues (51). Taken together, these observations suggest that susceptibility to infection may be the default state in some primary tissues. Moreover, susceptibility to FeLV may provide a useful marker to aid in the identification of further factors that confer tissue-specific susceptibility to genome instability in cancer as well as resistance to incoming retroviruses.
Can the risk of zoonotic infection with FeLV be discounted? Despite control measures, including vaccination, that have reduced the prevalence of FeLV in some parts of the world, this virus family remains one of the most important pathogens of cat species worldwide (52). Moreover, the limited barriers protecting human cells against FeLV infection contrast with the plethora of restriction factors that protect against the cross-species transmission of primate lentiviruses, where recent jumps have occurred (53). The successful transfer of other gammaretroviruses that use PIT-1 as an entry receptor from rodents to primates (5) argues that future adaptation of FeLV for zoonotic spread is not beyond the repertoire of this virus family. The possibility of human exposure to high doses of FeLV analogous to those employed in recent primate challenge experiments with XMRV (51, 54) is extremely remote; however, it may be interesting to revisit the apparent lack of adaptive immune responses in exposed individuals, e.g., veterinary workers (10), using more-sensitive techniques.

MATERIALS AND METHODS
Virus stocks. FeLV-B stocks were generated by transfection of an infectious molecular clone (pFGB) (55) and propagation in the feline fibroblast cell line AH927 or HEK human embryonic kidney cells as described previously (56). For FeLV-A stocks, HEK293 cells were transfected with pFGA-5 (20) to avoid possible recombination and generation of FeLV-B (57). FeLV infectivity was determined by titration on S ϩ L Ϫ QN10 cells (33).
Cells. Cell lines were derived from curated frozen stocks maintained in our laboratory (Jurkat, K562, CEM, Reh, Kasumi, U937, MCF-7) or by colleagues in the Glasgow Cancer Research UK Centre, including A. Biankin (pancreatic cancer TKCC series) and T. Holyoake (chronic myeloid leukemia lines KYO-1, BV173, KU812, LAMA84, and EM-2; B-cell acute lymphoblastic leukemia lines ALL/MIK and TOM-1). IMR90 fibroblasts were provided by P. Adams (48), while EBV-transformed lymphoblastoid cell lines (LCLs) were obtained from R. Jarrett, and HaCAT keratinocytes were provided by S. Graham. CEMSS cells were kindly provided by M. Malim (King's College, London, United Kingdom). CD34 ϩ cells were purified by positive selection using anti-CD34-conjugated magnetic microbeads (catalog no. 130-046-703; Miltenyi Biotec Inc.) from cord blood samples generously provided by the Children's Cancer and Leukemia Group (CCLG). Primary blood mononuclear cells were obtained from Cambridge Bioscience (1 sample) and the Scottish National Blood Transfusion Service (3 samples).
Dilution/spreading assay and Southern blot analysis. A total of 4 ϫ 10 5 KYO-1 cells or Reh cells or 5 ϫ 10 4 HEK293 cells were plated per well in 6 wells of a 12-well plate the day before infection. FeLV-B was harvested from subconfluent virus-producing HEK293 cells, filtered through 0.45-m filters, and serially diluted 1:10 in RPMI 1640 medium with 10% fetal calf serum (FCS). A 0.5-ml volume of filtered diluted virus was added per well, or medium alone was added to the control well, for 2 h. Cells were then fed with fresh medium and were expanded as normal for 14 days. Genomic DNA was extracted from phosphate-buffered saline (PBS)-washed, pelleted cells using a Qiagen DNeasy blood and tissue kit.
Quantitative real-time RT-PCR. RNA was extracted from cultured cells using a Qiagen RNeasy minikit, and cDNA was synthesized from 1 g RNA using a Qiagen QuantiTect reverse transcription kit. Aliquots (12.5 ng) of cDNA were amplified in triplicate using Power SYBR green PCR master mix (Thermo Fisher Scientific) and Qiagen QuantiTect primer assays for human hypoxanthine phosphoribosyltransferase (HPRT) (QT00059066), APOBEC3G (QT00070770), SLC20A1 (PIT-1) (QT00028763), or SLC19A2 (THTR1) (QT00007847) on an ABI 7500 real-time PCR system. Quantification was carried out relative to HEK293 cells, and values were normalized to HPRT expression.
Western blotting. PBS-washed, pelleted cells were lysed in whole-cell lysis buffer and protein concentrations measured by a Bio-Rad protein assay. Virion lysates were prepared from 2 ml precleared virion supernatant by centrifugation in a Beckman TL100 benchtop ultracentrifuge for 2 h at 37,000 rpm and 4°C. Virion pellets were resuspended in 75 l lysis buffer, and an equal volume (13 l) of virion protein or 10-g aliquots of cellular protein were separated on 4-to-12% NuPAGE Novex Bis-Tris gels in NuPAGE morpholinepropanesulfonic acid (MOPS) SDS buffer and were transferred to Hybond ECL membranes (Amersham) in NuPAGE transfer buffer. Following overnight blocking in 5% milk in Trisbuffered saline-Tween 20 (TBST), the filter was exposed to a 1:500 dilution of an anti-p27 Gag monoclonal antibody (VPG19.1; a gift of Brian Willett), and detection was carried out with the ECL detection reagent (GE Healthcare).

ACKNOWLEDGMENTS
We thank our colleagues at the Glasgow Cancer Centre for the provision of cell lines (Peter Adams, Andrew Biankin, Karen Blyth, Ruth Jarrett, Tessa Holyoake, and Kim Moran-Jones) and the Bloodwise Childhood Leukemia Cell Bank for cord blood samples. This work was supported by a joint program grant from Cancer Research UK (grant